Two examples do not describe a genus defined by what it binds, and a $1.2 billion verdict falls with the claims

2021-08-26Fed. Cir.

Juno Therapeutics, Inc. v. Kite Pharma, Inc., 10 F.4th 1330, No. 2020-1758 (Fed. Cir. Aug. 26, 2021) (precedential) (Moore, C.J., joined by Prost and O'Malley, JJ.), reversing the judgment of the United States District Court for the Central District of California, No. 2:17-cv-07639-PSG-KS (Gutierrez, J.)

United States Patent No. 7,446,190 claims a nucleic acid polymer encoding a three-part chimeric antigen receptor for a T cell: a CD3 zeta signaling domain, a CD28 costimulatory region, and a binding element that specifically interacts with a selected target. Claims 3 and 9 require the binding element to be a single-chain antibody variable fragment, an scFv, for any target, and claims 5 and 11 limit the target to CD19. The specification discloses two scFvs by name only, one derived from the SJ25C1 antibody that binds CD19 and one derived from J591 that binds PSMA, without their amino acid sequences, on a priority date of May 28, 2002. A jury found the claims not invalid, and the district court entered judgment for $1,200,322,551.50. The Federal Circuit reversed in an opinion by Chief Judge Moore. For genus claims using functional language, the written description must show that the applicant invented species sufficient to support the functionally defined genus, either by a representative number of species or by structural features common to the genus. The claims cover what Kite called millions of billions of possible scFvs, the patent gives nothing beyond alphanumeric designations to tell which scFvs bind which targets, only four or five CD19-specific scFvs were known at the priority date, and the inventor's subsequent work was irrelevant because written description is assessed as of the filing date.

What it changesMachine learning claims are drafted as functional genera all the time: a model trained to classify a defect, a network configured to predict a failure. This decision is the test those claims will face. A specification that reports one or two architectures and claims every model that achieves the result has described two species of an enormous genus, and if the field is unpredictable in the sense that the skilled person cannot tell from the disclosure which architectures and training regimes will achieve the function, the claim is unsupported no matter how many models the inventor builds afterward. Drafting should respond in two ways. Describe enough species, with distinct architectures, loss functions and data regimes actually run, with results, and state the structural relationship that makes them work, so the genus has a visible boundary. Then claim in tiers, with narrow species claims that survive if the genus falls, because the size of the verdict reversed here is the measure of what a functional-only claim set can lose.

More on Written description, enablement and claim scope · Litigation, damages and PTAB challenges

← All recent developments · Search all developments

Burbage Law is the practice of Jon Burbage, a registered patent attorney. 342 United States patents have issued from applications he wrote; the full list is on the Patents page. More about Jon Burbage.

This is general information about a published decision, agency guidance or piece of legislation. It is not legal advice and does not create an attorney-client relationship. Items concerning offices outside the United States are reported for information only and are not counsel on the law of those countries.